AI deployment costs spiral as enterprises struggle to track spending

Managing artificial intelligence expenses has become increasingly difficult for enterprises as the pace of agent deployment outstrips their capacity to observe and control spending, new findings from Splunk reveal. The Cisco subsidiary's State of Observability 2026 report surveyed 2,050 ITOps and engineering professionals to assess the challenge.

The survey uncovered troubling patterns around AI cost management. More than half of respondents indicated they lack the ability to govern AI spending, given that each prompt execution, model re-run and token consumption carries distinct pricing. Additionally, 60% of respondents reported that sudden spikes in AI-related charges create operational difficulties beyond their baseline cost expectations.

According to Mitch Ashley, VP and practice lead for CIO and technology buyers at The Futurum Group, the nature of AI expenses has fundamentally shifted. "The cost of AI has transitioned from a fixed line item to a variable cost that compounds at scale with little visibility," Ashley explained. He added that "a monthly report can't govern a cost that spikes in one afternoon."

Tracking and controlling variable costs

In response to unpredictable AI expenses, technology leaders are turning to specialized tools and methodologies to manage spending more effectively.

Ashley emphasized the importance of tools that provide granular visibility into spending patterns. "The most useful tools track spend back to its source and manage what's consumed at that point of origin," he said. "They tie tokens and calls to a specific agent, workflow, team, and trace, and enforce AI FinOps budget thresholds and rate limits at the agent operational layer before a request is sent."

Beyond instrumentation, Ashley recommended several operational practices for cost governance. Organizations should designate specific individuals responsible for AI workload expenses, establish named budgets for each agent or workflow from inception and monitor for cost deviations that warrant investigation.

Additional strategies include directing routine workloads toward smaller, less expensive models and incorporating consumption limits into vendor agreements.

Vendor solutions emerging

Technology vendors have begun addressing the AI cost management problem with new capabilities. Google introduced flexible billing arrangements and spending controls for AI agents in August, including monthly expenditure caps. Oracle deployed limited token bundles earlier in the year to align agentic pricing more closely with business outcomes, while AWS unveiled a preview of a FinOps agent designed to identify cost irregularities.

Observability challenges compound the issue

The complexity of AI cost management is being amplified by broader observability challenges. According to research from Enterprise Management Associates, enterprises are increasingly relying on observability tools to gain visibility into complex technology stacks, yet many organizations are struggling with tool proliferation. Three in four enterprise IT organizations deploy up to 12 different observability tools across cloud, infrastructure and network layers, a situation that itself contributes to rising AI expenses.

The emergence of AI agents is reshaping what observability tools must accomplish. Ashley noted that observability platforms need to provide insight into agent resource consumption, decision-making loops and system interactions. He emphasized that CIOs should scrutinize how observability vendors "price, sample and retain AI telemetry."

"AI workloads emit a trace for every prompt, tool call, and model response, so the observability bill climbs with the model bill," Ashley explained.